Related Experiment Video
Updated: Jan 7, 2026
07:25
Predicting Amputation using Local Circulating Mononuclear Progenitor Cells in Angioplasty-treated Patients with Critical Limb Ischemia
Published on: September 22, 2020
3.7K
A Study Protocol on Risk Prediction Modelling of Mortality and In-Hospital Major Bleeding Following Percutaneous
Mohammad Rocky Khan Chowdhury1,2, Mamunur Rashid3, Dion Stub1,4
1Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Melbourne, VIC 3004, Australia.
Methods and Protocols
|December 24, 2025
Summary
Machine learning models predict post-PCI outcomes like mortality and bleeding more accurately than regression. This study identifies key predictors to create new risk-scoring tools for better clinical decisions.
Area of Science:
- Cardiovascular Medicine
- Biostatistics
- Artificial Intelligence in Healthcare
Background:
- Percutaneous coronary intervention (PCI) is a common procedure with significant post-procedural risks.
- Existing risk prediction models often rely on traditional regression, which may not fully capture complex patient data.
- There is a need for more accurate and adaptable predictive tools for post-PCI outcomes.
Purpose of the Study:
- To develop and validate machine learning (ML)-based predictive models for 30-day mortality, in-hospital major bleeding, and one-year mortality after PCI.
- To identify key predictors of these adverse outcomes using advanced ML techniques.
- To create simplified, robust risk-scoring models for clinical decision support.
Main Methods:
- Analysis of a large dataset (104,665 cases) from the Victorian Cardiac Outcomes Registry (VCOR) (2013-2022).
- Application of ML algorithms, Boruta feature selection, multiple imputation for missing data, and SHapley Additive exPlanations (SHAP) for interpretability.
- Utilized 10-fold cross-validation, Adaptive Synthetic resampling for class imbalance, and external validation.
Main Results:
- Identification of influential predictors for 30-day mortality, in-hospital major bleeding, and one-year mortality post-PCI.
- Development of validated ML models demonstrating superior predictive performance compared to traditional methods.
- Conversion of key variables into simplified numeric scores for practical risk assessment.
Conclusions:
- Machine learning offers a powerful approach for predicting complex post-PCI outcomes.
- The identified key predictors and developed risk-scoring models can enhance clinical decision-making and patient risk stratification.
- This study provides a foundation for improved patient management and outcome prediction following PCI.
